## kernels.peek_kind()


The artifact type, read from the stream's schema message.


Usage

``` python
kernels.peek_kind(data)
```


What this saves is the DECODE, not the read. A router can dispatch a payload it is about to hand somewhere else without ever reconstructing the pandas frames and numpy arrays inside it - but it is not a cheap peek at a small prefix, and how little it reads depends entirely on the kind. For [PredictiveDistribution](kernels.PredictiveDistribution.md#ibnr.kernels.PredictiveDistribution), [MackFit](kernels.MackFit.md#ibnr.kernels.MackFit), [CDRResult](kernels.CDRResult.md#ibnr.kernels.CDRResult) and [Triangle](Triangle.md#ibnr.Triangle) the payload sits in record batches AFTER the schema, so this touches almost none of it. For `DataFrame`, `ForecastPanel` and `MackFitPanel` the body table is empty and the entire payload rides in the schema metadata as nested streams, so this reads 100% of the bytes - it simply does not unpack any of them.